摘要 |
PURPOSE: To improve reproduced sound quality without changing a bit rate by selecting a code vector in which an error vector is evaluated by a weighing coefficient vector and an visual error is made the minimum. CONSTITUTION: Error vectors obtained as differences of each vector component are calculated for all code vectors and cepstrum coefficient vector in an error vector calculating means 108. This error vector is converted to conversion error vector in a frequency region by a conversion means 109 performing discrete Fourier conversion. A weighing coefficient vector in which a visual characteristic is considered is also obtained as a function of a frequency based on spectrum amplitude information obtained by a spectrum amplitude information means 105 in a weighing coefficient vector calculating means 110. And in a representative vector selecting means 111, the sum total of the product of square of each element of weighing coefficient vectors and each element of conversion error vectors is calculated, a code vector in which the value is made the minimum is selected, and its index is encoded. |